In this notebook I will perform intent classification using three different models to see which one is better for this task. I will be using PyTorch. the data I used for this notebook is provided. The names are testlabel, testseq.in, train_label, trainseq.in
In this notebook you can expect to see:
- Uploading the data and understanding it
- Preprocessing the data and some feature engineering
- Creating my own tokenization algorithm
- Creating my own accuracy, confusion matrix and training functions
- Feeding the data into the models and performing statistical inference
The models I used for intent classification are:
- Simple ANN model
- BiDirectional LSTM model based off of Yu Wang, Yilin Shen & Hongxia Jin
- BERT transformer model (transfer learning). I used a pretrained model and changed it up a bit to work for my own problem.
I think Natural Language Processing is very interesting and hope to work on more similar projects.